Enregistré dans:
| Auteurs principaux: | , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2401.15178 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910313362751488 |
|---|---|
| author | Brown, Henry J. Grabovsky, Yury |
| author_facet | Brown, Henry J. Grabovsky, Yury |
| contents | The feasibility of extrapolation of completely monotone functions can be quantified by examining the worst case scenario, whereby a pair of completely monotone functions agree on a given interval to a given relative precision, but differ as much as it is theoretically possible at a given point. We show that extrapolation is impossible to the left of the interval, while the maximal discrepancy to the right exhibits a power law typical for extrapolation of similar classes of complex analytic functions. The power law exponent is derived explicitly, and shows a precipitous drop immediately beyond the right end-point, with a subsequent decay to zero inversely proportional to the distance from the interval. The local extrapolation problem, where the worst discrepancy from a given completely monotone function is sought, is also analyzed. In this case explicit and easily verifiable optimality conditions are derived, enabling us to solve the problem exactly for a single decaying exponential. In the general case, our approach leads to a natural algorithm for computing solutions to the local extrapolation problem numerically. The methods developed in this paper can easily be adapted to other classes of analytic functions represented as integral transforms of positive measures with analytic kernels. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_15178 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | On feasibility of extrapolation of completely monotone functions Brown, Henry J. Grabovsky, Yury Complex Variables Optimization and Control 49K40, 90C46, 30B40, 44A10, 26A48, 41A30, 41A05, 30A10 The feasibility of extrapolation of completely monotone functions can be quantified by examining the worst case scenario, whereby a pair of completely monotone functions agree on a given interval to a given relative precision, but differ as much as it is theoretically possible at a given point. We show that extrapolation is impossible to the left of the interval, while the maximal discrepancy to the right exhibits a power law typical for extrapolation of similar classes of complex analytic functions. The power law exponent is derived explicitly, and shows a precipitous drop immediately beyond the right end-point, with a subsequent decay to zero inversely proportional to the distance from the interval. The local extrapolation problem, where the worst discrepancy from a given completely monotone function is sought, is also analyzed. In this case explicit and easily verifiable optimality conditions are derived, enabling us to solve the problem exactly for a single decaying exponential. In the general case, our approach leads to a natural algorithm for computing solutions to the local extrapolation problem numerically. The methods developed in this paper can easily be adapted to other classes of analytic functions represented as integral transforms of positive measures with analytic kernels. |
| title | On feasibility of extrapolation of completely monotone functions |
| topic | Complex Variables Optimization and Control 49K40, 90C46, 30B40, 44A10, 26A48, 41A30, 41A05, 30A10 |
| url | https://arxiv.org/abs/2401.15178 |